• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
2<html>
3<head>
4
5<meta http-equiv="Content-Type" content="text/html; charset=iso-8859-15"/>
6<title>Ogg Documentation</title>
7
8<style type="text/css">
9body {
10  margin: 0 18px 0 18px;
11  padding-bottom: 30px;
12  font-family: Verdana, Arial, Helvetica, sans-serif;
13  color: #333333;
14  font-size: .8em;
15}
16
17a {
18  color: #3366cc;
19}
20
21img {
22  border: 0;
23}
24
25#xiphlogo {
26  margin: 30px 0 16px 0;
27}
28
29#content p {
30  line-height: 1.4;
31}
32
33h1, h1 a, h2, h2 a, h3, h3 a {
34  font-weight: bold;
35  color: #ff9900;
36  margin: 1.3em 0 8px 0;
37}
38
39h1 {
40  font-size: 1.3em;
41}
42
43h2 {
44  font-size: 1.2em;
45}
46
47h3 {
48  font-size: 1.1em;
49}
50
51li {
52  line-height: 1.4;
53}
54
55#copyright {
56  margin-top: 30px;
57  line-height: 1.5em;
58  text-align: center;
59  font-size: .8em;
60  color: #888888;
61  clear: both;
62}
63</style>
64
65</head>
66
67<body>
68
69<div id="xiphlogo">
70  <a href="http://www.xiph.org/"><img src="fish_xiph_org.png" alt="Fish Logo and Xiph.org"/></a>
71</div>
72
73<h1>Ogg logical bitstream framing</h1>
74
75<h2>Ogg bitstreams</h2>
76
77<p>The Ogg transport bitstream is designed to provide framing, error
78protection and seeking structure for higher-level codec streams that
79consist of raw, unencapsulated data packets, such as the Vorbis audio
80codec or Theora video codec.</p>
81
82<h2>Application example: Vorbis</h2>
83
84<p>Vorbis encodes short-time blocks of PCM data into raw packets of
85bit-packed data. These raw packets may be used directly by transport
86mechanisms that provide their own framing and packet-separation
87mechanisms (such as UDP datagrams). For stream based storage (such as
88files) and transport (such as TCP streams or pipes), Vorbis uses the
89Ogg bitstream format to provide framing/sync, sync recapture
90after error, landmarks during seeking, and enough information to
91properly separate data back into packets at the original packet
92boundaries without relying on decoding to find packet boundaries.</p>
93
94<h2>Design constraints for Ogg bitstreams</h2>
95
96<ol>
97<li>True streaming; we must not need to seek to build a 100%
98  complete bitstream.</li>
99<li>Use no more than approximately 1-2% of bitstream bandwidth for
100  packet boundary marking, high-level framing, sync and seeking.</li>
101<li>Specification of absolute position within the original sample
102  stream.</li>
103<li>Simple mechanism to ease limited editing, such as a simplified
104  concatenation mechanism.</li>
105<li>Detection of corruption, recapture after error and direct, random
106  access to data at arbitrary positions in the bitstream.</li>
107</ol>
108
109<h2>Logical and Physical Bitstreams</h2>
110
111<p>A <em>logical</em> Ogg bitstream is a contiguous stream of
112sequential pages belonging only to the logical bitstream. A
113<em>physical</em> Ogg bitstream is constructed from one or more
114than one logical Ogg bitstream (the simplest physical bitstream
115is simply a single logical bitstream). We describe below the exact
116formatting of an Ogg logical bitstream. Combining logical
117bitstreams into more complex physical bitstreams is described in the
118<a href="oggstream.html">Ogg bitstream overview</a>. The exact
119mapping of raw Vorbis packets into a valid Ogg Vorbis physical
120bitstream is described in the Vorbis I Specification.</p>
121
122<h2>Bitstream structure</h2>
123
124<p>An Ogg stream is structured by dividing incoming packets into
125segments of up to 255 bytes and then wrapping a group of contiguous
126packet segments into a variable length page preceded by a page
127header. Both the header size and page size are variable; the page
128header contains sizing information and checksum data to determine
129header/page size and data integrity.</p>
130
131<p>The bitstream is captured (or recaptured) by looking for the beginning
132of a page, specifically the capture pattern. Once the capture pattern
133is found, the decoder verifies page sync and integrity by computing
134and comparing the checksum. At that point, the decoder can extract the
135packets themselves.</p>
136
137<h3>Packet segmentation</h3>
138
139<p>Packets are logically divided into multiple segments before encoding
140into a page. Note that the segmentation and fragmentation process is a
141logical one; it's used to compute page header values and the original
142page data need not be disturbed, even when a packet spans page
143boundaries.</p>
144
145<p>The raw packet is logically divided into [n] 255 byte segments and a
146last fractional segment of &lt; 255 bytes. A packet size may well
147consist only of the trailing fractional segment, and a fractional
148segment may be zero length. These values, called "lacing values" are
149then saved and placed into the header segment table.</p>
150
151<p>An example should make the basic concept clear:</p>
152
153<pre>
154<tt>
155raw packet:
156  ___________________________________________
157 |______________packet data__________________| 753 bytes
158
159lacing values for page header segment table: 255,255,243
160</tt>
161</pre>
162
163<p>We simply add the lacing values for the total size; the last lacing
164value for a packet is always the value that is less than 255. Note
165that this encoding both avoids imposing a maximum packet size as well
166as imposing minimum overhead on small packets (as opposed to, eg,
167simply using two bytes at the head of every packet and having a max
168packet size of 32k. Small packets (&lt;255, the typical case) are
169penalized with twice the segmentation overhead). Using the lacing
170values as suggested, small packets see the minimum possible
171byte-aligned overhead (1 byte) and large packets, over 512 bytes or
172so, see a fairly constant ~.5% overhead on encoding space.</p>
173
174<p>Note that a lacing value of 255 implies that a second lacing value
175follows in the packet, and a value of &lt; 255 marks the end of the
176packet after that many additional bytes. A packet of 255 bytes (or a
177multiple of 255 bytes) is terminated by a lacing value of 0:</p>
178
179<pre><tt>
180raw packet:
181  _______________________________
182 |________packet data____________|          255 bytes
183
184lacing values: 255, 0
185</tt></pre>
186
187<p>Note also that a 'nil' (zero length) packet is not an error; it
188consists of nothing more than a lacing value of zero in the header.</p>
189
190<h3>Packets spanning pages</h3>
191
192<p>Packets are not restricted to beginning and ending within a page,
193although individual segments are, by definition, required to do so.
194Packets are not restricted to a maximum size, although excessively
195large packets in the data stream are discouraged; the Ogg
196bitstream specification strongly recommends nominal page size of
197approximately 4-8kB (large packets are foreseen as being useful for
198initialization data at the beginning of a logical bitstream).</p>
199
200<p>After segmenting a packet, the encoder may decide not to place all the
201resulting segments into the current page; to do so, the encoder places
202the lacing values of the segments it wishes to belong to the current
203page into the current segment table, then finishes the page. The next
204page is begun with the first value in the segment table belonging to
205the next packet segment, thus continuing the packet (data in the
206packet body must also correspond properly to the lacing values in the
207spanned pages. The segment data in the first packet corresponding to
208the lacing values of the first page belong in that page; packet
209segments listed in the segment table of the following page must begin
210the page body of the subsequent page).</p>
211
212<p>The last mechanic to spanning a page boundary is to set the header
213flag in the new page to indicate that the first lacing value in the
214segment table continues rather than begins a packet; a header flag of
2150x01 is set to indicate a continued packet. Although mandatory, it
216is not actually algorithmically necessary; one could inspect the
217preceding segment table to determine if the packet is new or
218continued. Adding the information to the packet_header flag allows a
219simpler design (with no overhead) that needs only inspect the current
220page header after frame capture. This also allows faster error
221recovery in the event that the packet originates in a corrupt
222preceding page, implying that the previous page's segment table
223cannot be trusted.</p>
224
225<p>Note that a packet can span an arbitrary number of pages; the above
226spanning process is repeated for each spanned page boundary. Also a
227'zero termination' on a packet size that is an even multiple of 255
228must appear even if the lacing value appears in the next page as a
229zero-length continuation of the current packet. The header flag
230should be set to 0x01 to indicate that the packet spanned, even though
231the span is a nil case as far as data is concerned.</p>
232
233<p>The encoding looks odd, but is properly optimized for speed and the
234expected case of the majority of packets being between 50 and 200
235bytes (note that it is designed such that packets of wildly different
236sizes can be handled within the model; placing packet size
237restrictions on the encoder would have only slightly simplified design
238in page generation and increased overall encoder complexity).</p>
239
240<p>The main point behind tracking individual packets (and packet
241segments) is to allow more flexible encoding tricks that requiring
242explicit knowledge of packet size. An example is simple bandwidth
243limiting, implemented by simply truncating packets in the nominal case
244if the packet is arranged so that the least sensitive portion of the
245data comes last.</p>
246
247<h3>Page header</h3>
248
249<p>The headering mechanism is designed to avoid copying and re-assembly
250of the packet data (ie, making the packet segmentation process a
251logical one); the header can be generated directly from incoming
252packet data. The encoder buffers packet data until it finishes a
253complete page at which point it writes the header followed by the
254buffered packet segments.</p>
255
256<h4>capture_pattern</h4>
257
258<p>A header begins with a capture pattern that simplifies identifying
259pages; once the decoder has found the capture pattern it can do a more
260intensive job of verifying that it has in fact found a page boundary
261(as opposed to an inadvertent coincidence in the byte stream).</p>
262
263<pre><tt>
264 byte value
265
266  0  0x4f 'O'
267  1  0x67 'g'
268  2  0x67 'g'
269  3  0x53 'S'
270</tt></pre>
271
272<h4>stream_structure_version</h4>
273
274<p>The capture pattern is followed by the stream structure revision:</p>
275
276<pre><tt>
277 byte value
278
279  4  0x00
280</tt></pre>
281
282<h4>header_type_flag</h4>
283
284<p>The header type flag identifies this page's context in the bitstream:</p>
285
286<pre><tt>
287 byte value
288
289  5  bitflags: 0x01: unset = fresh packet
290	               set = continued packet
291	       0x02: unset = not first page of logical bitstream
292                       set = first page of logical bitstream (bos)
293	       0x04: unset = not last page of logical bitstream
294                       set = last page of logical bitstream (eos)
295</tt></pre>
296
297<h4>absolute granule position</h4>
298
299<p>(This is packed in the same way the rest of Ogg data is packed; LSb
300of LSB first. Note that the 'position' data specifies a 'sample'
301number (eg, in a CD quality sample is four octets, 16 bits for left
302and 16 bits for right; in video it would likely be the frame number.
303It is up to the specific codec in use to define the semantic meaning
304of the granule position value). The position specified is the total
305samples encoded after including all packets finished on this page
306(packets begun on this page but continuing on to the next page do not
307count). The rationale here is that the position specified in the
308frame header of the last page tells how long the data coded by the
309bitstream is. A truncated stream will still return the proper number
310of samples that can be decoded fully.</p>
311
312<p>A special value of '-1' (in two's complement) indicates that no packets
313finish on this page.</p>
314
315<pre><tt>
316 byte value
317
318  6  0xXX LSB
319  7  0xXX
320  8  0xXX
321  9  0xXX
322 10  0xXX
323 11  0xXX
324 12  0xXX
325 13  0xXX MSB
326</tt></pre>
327
328<h4>stream serial number</h4>
329
330<p>Ogg allows for separate logical bitstreams to be mixed at page
331granularity in a physical bitstream. The most common case would be
332sequential arrangement, but it is possible to interleave pages for
333two separate bitstreams to be decoded concurrently. The serial
334number is the means by which pages physical pages are associated with
335a particular logical stream. Each logical stream must have a unique
336serial number within a physical stream:</p>
337
338<pre><tt>
339 byte value
340
341 14  0xXX LSB
342 15  0xXX
343 16  0xXX
344 17  0xXX MSB
345</tt></pre>
346
347<h4>page sequence no</h4>
348
349<p>Page counter; lets us know if a page is lost (useful where packets
350span page boundaries).</p>
351
352<pre><tt>
353 byte value
354
355 18  0xXX LSB
356 19  0xXX
357 20  0xXX
358 21  0xXX MSB
359</tt></pre>
360
361<h4>page checksum</h4>
362
363<p>32 bit CRC value (direct algorithm, initial val and final XOR = 0,
364generator polynomial=0x04c11db7). The value is computed over the
365entire header (with the CRC field in the header set to zero) and then
366continued over the page. The CRC field is then filled with the
367computed value.</p>
368
369<p>(A thorough discussion of CRC algorithms can be found in <a
370href="http://www.ross.net/crc/download/crc_v3.txt">"A
371Painless Guide to CRC Error Detection Algorithms"</a> by Ross
372Williams <a href="mailto:ross@ross.net">ross@ross.net</a>.)</p>
373
374<pre><tt>
375 byte value
376
377 22  0xXX LSB
378 23  0xXX
379 24  0xXX
380 25  0xXX MSB
381</tt></pre>
382
383<h4>page_segments</h4>
384
385<p>The number of segment entries to appear in the segment table. The
386maximum number of 255 segments (255 bytes each) sets the maximum
387possible physical page size at 65307 bytes or just under 64kB (thus
388we know that a header corrupted so as destroy sizing/alignment
389information will not cause a runaway bitstream. We'll read in the
390page according to the corrupted size information that's guaranteed to
391be a reasonable size regardless, notice the checksum mismatch, drop
392sync and then look for recapture).</p>
393
394<pre><tt>
395 byte value
396
397 26 0x00-0xff (0-255)
398</tt></pre>
399
400<h4>segment_table (containing packet lacing values)</h4>
401
402<p>The lacing values for each packet segment physically appearing in
403this page are listed in contiguous order.</p>
404
405<pre><tt>
406 byte value
407
408 27 0x00-0xff (0-255)
409 [...]
410 n  0x00-0xff (0-255, n=page_segments+26)
411</tt></pre>
412
413<p>Total page size is calculated directly from the known header size and
414lacing values in the segment table. Packet data segments follow
415immediately after the header.</p>
416
417<p>Page headers typically impose a flat .25-.5% space overhead assuming
418nominal ~8k page sizes. The segmentation table needed for exact
419packet recovery in the streaming layer adds approximately .5-1%
420nominal assuming expected encoder behavior in the 44.1kHz, 128kbps
421stereo encodings.</p>
422
423<div id="copyright">
424  The Xiph Fish Logo is a
425  trademark (&trade;) of Xiph.Org.<br/>
426
427  These pages &copy; 1994 - 2005 Xiph.Org. All rights reserved.
428</div>
429
430</body>
431</html>
432